ASSESSMENT OF DEEP LEARNING TECHNIQUES IN SENTIMENT EVALUATION FROM TWITTER STATISTICS

Abstract

This take a look at provides a assessment of different deep gaining knowledge of methods used for sentiment evaluation in Twitter statistics. on this domain, deep learning (DL) strategies, which make a contribution on the equal time to the solution of a extensive variety of issues, won popularity amongst researchers. specifically, two classes of neural networks are utilized, convolution neural networks (CNN), which are particularly performant in the location of photograph processing and recurrent neural networks (RNN) which might be implemented with success in natural language processing (NLP) duties. on this paintings we compare and compare ensembles and combinations of CNN and a category of RNN the lengthy shortterm memory (LSTM) networks. moreover, we compare one of a kind phrase embedding systems including the Word2Vec and the worldwide vectors for phrase representation (GloVe) fashions. For the evaluation of those strategies we used information furnished by way of the global workshop on semantic assessment (SemEval), that's one of the most famous international workshops at the location. Diverse tests and combos are applied and best scoring values for every version are as compared in terms of their overall performance. This take a look at contributes to the sphere of sentiment analysis with the aid of analyzing the performances, blessings and barriers of the above methods with an evaluation method underneath a unmarried testing framework with the identical dataset and computing surroundings.